December 3, 2025

OpenAI has trained its LLM to confess to bad behavior

OpenAI is testing another new way to expose the complicated processes at work inside large language models. Researchers at the company can make an LLM produce what they call a confession, in which the model explains how it carried out a task and (most of the time) owns up to any bad behavior. Figuring out…

TopTechTrends

Nominations are now open for our global 2026 Innovators Under 35 competition

December 1, 2025
We have some exciting news: Nominations are now open for MIT Technology Review’s 2026 Innovators Under 35 competition. This annual list recognizes 35 of the world’s best young scientists and inventors, and our newsroom has produced it for more than two decades.  It’s free to nominate yourself or someone you know, and it only takes…

An AI model trained on prison phone calls now looks for planned crimes in those calls

December 1, 2025
A US telecom company trained an AI model on years of inmates’ phone and video calls and is now piloting that model to scan their calls, texts, and emails in the hope of predicting and preventing crimes.  Securus Technologies president Kevin Elder told MIT Technology Review that the company began building its AI tools in…

Airbus orders software fix to thousands of planes due to solar radiation risk

November 29, 2025
Flights were delayed and cancelled globally after Airbus ordered fixes to 6,000 of its A320 series planes, according to The Guardian. The company said it’s taking action because “analysis of a recent event involving an A320 Family aircraft has revealed that intense solar radiation may corrupt data critical to the functioning of flight controls.” Citing […]